docs-graph

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • Command Execution (MEDIUM): The skill uses the Bash tool to execute python3 scripts/graph.py for graph generation, querying, and verification. This constitutes execution of code not contained within the analyzed skill file.
  • Auto-Execution / Persistence (MEDIUM): The documentation promotes the use of PostToolUse hooks in .claude/settings.local.json to automatically trigger the Python script whenever files in the plans/ directory are modified. This creates a persistent trigger for code execution based on file system events.
  • Indirect Prompt Injection (LOW): The skill processes untrusted markdown content from the plans/ directory to build its knowledge graph.
  • Ingestion points: All markdown files within the plans/ directory and its subdirectories.
  • Boundary markers: None identified; the skill parses raw file content for [[wikilinks]] patterns.
  • Capability inventory: Execution of local Python scripts via python3, file system reading via Read, and file discovery via Glob.
  • Sanitization: No evidence of sanitization or escaping for data interpolated into the Mermaid diagrams or console output.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:32 PM